Leakage detection in pipelines is a critical concern in captive power plants, where
undetected leaks can lead to material wastage, environmental hazards, and unexpected
plant shutdowns. Traditional inspection methods are often manual, time-consuming, and
incapable of providing timely alerts. This work presents an IoT-based intelligent leakage
detection and monitoring system specifically designed for pipelines in captive power plants,
focusing on fly ash slurry and similar industrial flows.
The proposed system integrates multiple electromagnetic flow sensors positioned at
strategic locations along the pipeline network to continuously measure flow rates. These
sensors are interfaced with ESP32 microcontrollers and ADS1115 ADC modules to ensure
accurate data acquisition. Using a SIM A7672S LTE module, the system transmits
data to a cloud server in real time. A web-based dashboard displays live flow values
through intuitive gauge and line charts, offering operators immediate visibility into pipeline
performance. Beyond simple threshold-based alerts, the system employs a machine learning
approach using a One-Class Support Vector Machine (OCSVM). The model is trained
on historical data representing normal and leak conditions, leveraging statistical features
such as moving averages, overlapping window means, and covariance matrices to detect
anomalies. When the system identifies deviations from learned patterns, it promptly raises
alerts, enabling early intervention before minor leaks escalate into major failures.Field trials
conducted in an operational captive power plant demonstrated the system&rsquos capability to
detect subtle leakages that might escape manual checks. Its modular IoT architecture ensures
scalability, allowing integration with additional parameters like pressure and temperature
for comprehensive health monitoring. The wireless design facilitates deployment in remote
or difficult-to-access sections of the plant. Overall, this intelligent leakage detection system
shifts maintenance from a reactive to a predictive paradigm, reducing downtime, minimizing
environmental risks, and achieving substantial cost savings. Future enhancements will focus
on integrating advanced deep learning models and redundant communication paths to ensure
reliability even during network failures. This work underscores the potential of IoT and
AI to revolutionize industrial pipeline monitoring, driving safer and more efficient plant
operations.